HIGHER-ORDER DIMENSIONS OF THE BIG FIVE PERSONALITY TRAITS AND THE BIG SIX VOCATIONAL INTEREST TYPES

PERSONNEL PSYCHOLOGY2005, 58, 447–478

HIGHER-ORDER DIMENSIONS OF THE BIG FIVEPERSONALITY TRAITS AND THE BIG SIXVOCATIONAL INTEREST TYPES

MICHAEL K. MOUNT, MURRAY R. BARRICKHenry B. Tippie College of Business

University of Iowa

STEVE M. SCULLENNorth Carolina State University

JAMES ROUNDSUniversity of Illinois

The purpose of this study was to identify higher-order dimensionsthat explain the relationships among the Big 6 interest types and theBig 5 personality traits. Meta-analyses were conducted to identifyan 11 × 11 true score correlation matrix of interest and personal-ity attributes. Cluster analysis and nonmetric multidimensional scalingwere used to identify 3 dimensions that explained relations among the11 attributes: (a) Interests versus Personality Traits; (b) Striving for Ac-complishment Versus Striving for Personal Growth, and (c) Interactingwith People Versus Interacting with Things. Overall, results clarified therelationships among interests and personality traits by showing that 3rather than 2 dimensions best explain the relationships among interestsand personality traits.

Personality traits and vocational interests are two major, noncogni-tive individual difference domains in the field of psychology. Both setsof dispositional attributes are important because they influence numer-ous outcomes associated with work and life success. One common threadthat links personality traits and vocational interests is that they influencebehavior through motivational processes. That is, they influence choicesindividuals make about which tasks and activities to engage in, how mucheffort to exert on those tasks, and how long to persist with those tasks.Although psychologists have conducted hundreds of studies that investi-gate one or both topics, the precise nature of the linkages between the twodomains remains ambiguous and controversial.

The authors would like to thank Paul Muchinsky, Robert Hogan, and John Holland forhelpful comments on an earlier version of this manuscript.

Paul Sackett served as guest editor for this article.Correspondence and requests for reprints should be addressed to Michael K. Mount,

University of Iowa, Henry B. Tippie College of Business, Iowa City, Iowa 52244-1000,319-335-0953; [email protected].

COPYRIGHT C© 2005 BLACKWELL PUBLISHING, INC.

447

448 PERSONNEL PSYCHOLOGY

The purpose of the present study is to investigate the nomological netof the relationships between attributes that comprise the two domains.Cronbach and Meehl (1955) developed the idea of the nomological, orlawful, network in order to provide construct validity evidence of mea-sures. We seek to identify the higher-order dimensions that explain therelationships between vocational interests, as represented by the Big Sixinterest types (e.g., Holland, 1985), and personality traits, as representedby the Big Five personality dimensions (e.g., Digman, 1990). Identify-ing the lawful relationships that underlie these two individual differencedomains has important practical and theoretical implications. In particu-lar, we can gain a more comprehensive understanding of the fundamentalnature of motivation by considering the two domains jointly, rather thanstudying discrete personality traits or interest types in isolation. In turn,understanding the nature of motivation should lead to the developmentof better theories and measures that help explain behavior that is undervolitional control.

Personality Traits—Definition and Structure

Definitions

There are many definitions of personality. (See Saucier and Goldberg[2003] for a detailed discussion.) In fact, Allport (1937) called personal-ity one of the most abstract words in our language and listed 50 distinctmeanings that were derived from fields as diverse as theology, philoso-phy, sociology, law, and psychology. Although there is some disagree-ment among contemporary personality theorists about the meaning ofpersonality, there is agreement that what people do is influenced by sta-ble characteristics, that is, their personality. Hogan (1991) believes thatpersonality has two different meanings and the failure to separate themleads to confusion. The first is a person’s social reputation, which refersto the way an individual is perceived by others; it is personality from theobserver’s perspective and is public and verifiable. The second refers tothe structures, dynamics, processes, and propensities that explain why aperson behaves in a characteristic way; it is private and must be inferred.McCrae and Costa (1989) define personality as enduring emotional, in-terpersonal, experiential, attitudinal, and motivational styles that explainbehavior in different situations. Funder (2001) defines personality as “anindividual’s characteristic pattern of thought, emotion, and behavior, to-gether with the psychological mechanisms—hidden or not—behind thosepatterns.” In sum, personality traits refer to the characteristics that arestable over time, provide the reasons for the person’s behavior, and are

MICHAEL K. MOUNT ET AL. 449

psychological in nature. They reflect who we are and in aggregate deter-mine our affective, behavioral, and cognitive style.

Five Factor Model Personality Dimensions

The five-factor model (FFM) of personality describes the basicdimensions of the normal personality. Generally speaking, there iswidespread agreement about the five personality dimensions and theircontent. (See Block [1995] for a contrarian view.) These dimensions(and prototypical characteristics) include: Extraversion (sociable, active,energetic), Agreeableness (cooperative, considerate, trusting), Conscien-tiousness (dependable, organized, persistent), Emotional Stability (calm,secure, unemotional), and Openness to Experience (imaginative, intellec-tual, artistically sensitive).

Vocational Interests—Definition and Structure

Definition

Researchers have provided numerous definitions of vocational inter-ests over the past 50 years. (See Dawis [1991] and Savickas [1999] fora comprehensive discussion of these definitions.) For example, Strong(1960) defined interests as activities that are liked and disliked. Similarly,Kuder (1977) viewed interests as being manifested in preferences for ac-tivities. Others saw interests as constellations of likes and dislikes thatlead to consistent patterns of behaviors (Berdie, 1944; Cole & Hanson,1978). Although these definitions are quite different from those providedfor personality, other definitions suggest that interests and personality arevery similar. Carter (1944) saw interests as based on evaluative attitudesand as such reflect stable personality traits. Moreover, Holland (1973,p. 7) stated the following: “In short, what we have called ‘vocational in-terests’ are simply another aspect of personality . . . If vocational interestsare expressions of personality, then it follows that interest inventories arepersonality inventories.” (See Holland [1999] for a detailed discussion ofthis issue.) Further, Bingham (1937) refers to an interest as a dispositionaltendency, which has affective, behavioral, and cognitive components, withdimensions of intensity and duration. This definition is also very similarto the definition of personality. Taken together, interests reflect long-termdispositional traits inherent in the person that influence behavior primarilythrough one’s preferences for certain environments, activities, and typesof people.


RIASEC Types

Holland’s (1959, 1966, 1973, 1997) theory of vocational personalitiesand work environments states that an employee’s satisfaction with a jobas well as propensity to leave that job, depend on the degree to whichhis/her personality matches the occupational environment. Although notwithout its critics (e.g., Gati, 1991), Holland’s (1997) RIASEC typologyhas been widely studied in the vocational literature. The RIASEC typol-ogy has emerged repeatedly in large samples (Rounds & Tracey, 1993;Tracey & Rounds, 1993), and its generalizability has been supported withcross-cultural analyses (Day & Rounds, 1997). Realistic people prefer ac-tivities involving the systematic manipulation of machinery, tools, or ani-mals. Investigative individuals tend to be analytical, curious, methodical,and precise. Artistic people tend to be expressive, nonconforming, origi-nal, and introspective. Individuals who are social enjoy working with andhelping others. Enterprising individuals enjoy those activities that entailinfluencing others to attain organizational goals or economic gain. Finally,conventional individuals enjoy the systematic manipulation of data, filingrecords, or reproducing materials.

Several studies using Holland’s RIASEC typology have shown thatcongruence between interests and environment is associated with greatersatisfaction (e.g., Assouline & Meir, 1987; Mount & Muchinsky, 1978;Spokane, 1985). As this suggests, vocational interests are broad, mul-tifaceted orientations that are associated with behaviors that reflect anindividual’s choice to engage in tasks and activities they like and to bein environments where they are surrounded by people who are similar tothem. Furthermore, the outcome that is central to researchers and the onemost likely influenced by these choices is the degree of satisfaction that re-sults from the congruence between interests and the work environments.

Hypotheses

Dimensions Within RIASEC Types and Within FFM Traits

A useful starting point for exploring the possible higher-order dimen-sions that are common across the two domains is to examine the evidencepertaining to higher-order dimensions within each domain. Three concep-tualizations of the higher-order dimensions of vocational interests havebeen discussed most prominently. Hogan (Hogan, 1983; Hogan & Blake,1999) argues that the RIASEC vocational interests consist of two funda-mental, bipolar dimensions, Conformity and Sociability. These two higher-order dimensions are centered at right angles to each other on the RIASECcircumplex. The Conformity dimension is aligned with the Conventional


type and Artistic type, and the Sociability dimension bisects two setsof interest types, ranging from Enterprising-Social types to Realistic-Investigative types. Prediger (Prediger, 1982; Prediger & Vansickle, 1992)proposed that interests and environments could be characterized by twosomewhat different bipolar dimensions. These two higher-order dimen-sions are centered on the RIASEC circumplex 30 degrees counterclock-wise from Hogan’s dimensions. The People/Things dimension character-izes differences between Social and Realistic types on opposite sides ofthe hexagon. The Data/Ideas dimension is defined by Enterprising andConventional versus Investigative and Artistic types, respectively. Gati(1991) found support for a three-group partition model of the structure ofinterests whereby RIASEC types collapse into three clusters ([Realistic,Investigative], [Artistic, Social], and [Enterprising, Conventional]).

Regarding the structure and dimensionality of personality, Saucier andGoldberg (2003) state that although the Big Five factor solution is robust,other meaningful factor structures also emerge. A higher-order structurethat has been supported empirically in several studies is a two-factor so-lution, which is related to positively valued, dynamic qualities associ-ated with socialization and individual ascendancy. These two factors havebeen alternatively labeled Factors α and β (Digman, 1997), Social Interestand Superiority Striving (Adler, 1939), Communion and Agency (Bakan,1966), Maintaining Popularity and Gaining Status (Hogan, 1982), andUnion and Individuation (Rank, 1945).

The Digman (1997) study is particularly informative because it wasderived from correlations based on 14 samples that assessed personal-ity using the FFM model personality instruments. He found support fortwo—and only two—higher-order factors. Factor α consisted of the FFMfactors of Conscientiousness, Emotional Stability, and Agreeableness,whereas Factor β consisted of the FFM factors of Extraversion and Open-ness to Experience. According to Digman (1997) Factor α pertains toimpulse restraint, conscience, low hostility and aggression, and neuroticdefense. Factor β refers to actualization of the self, venturesome encoun-ters with life, openness to new experiences, and use of one’s intellect. Weexamine whether the higher-order structure of a set of attributes that con-tains both personality and vocational interests is consistent with Digman’s(1997) Factors α and β, which were extracted from the FFM variablesalone.

Higher-Order Dimensions Between RIASEC and FFM Types

The fact that the six RIASEC types and five FFM traits representnoncognitive dispositional attributes that influence recurring patterns ofbehavior suggests that they are similar. However, these two sets of


dispositional attributes also differ fundamentally both in their relation-ship to motivational processes and their relationship to outcomes (e.g.,preferences versus implementation of intentions). The fundamental na-ture of this difference revolves around the fact that interests pertain toindividuals’ preferences, which influence outcomes associated with sat-isfaction with choices made. On the other hand, personality traits pertainto self-regulatory and motivational processes, which influence outcomesassociated with performance (or satisfactoriness) on chosen tasks. Con-sistent with these differences, Super (1973), Dawis (1980), and Hoganand Blake (1999) have discussed an underlying structure of these motiva-tional constructs that differ fundamentally in the ways of dealing with theenvironment. Interests are regarded as driving preferences for the typesof environments individuals like to work in. On the other hand, personal-ity traits determine the ways of interacting in those environments and arelikely to occupy a position at the opposite end on a dimension representingthese motivational constructs.

Although the nature and magnitude of the relations between per-sonality traits and interests have been systematically examined in nu-merous studies, there are four that are especially relevant to this study.Ackerman and Heggestad (1997) evaluated relations among personalitytraits, vocational interests, and intellectual abilities. They conducted anextensive meta-analysis of personality–intellectual ability correlations,reviewed literature on interest–intellectual ability correlations, and alsogauged the overlap between personality and interests based on three di-verse studies (but did not conduct a meta-analysis of personality–interestrelationships). With respect to the personality–interest correlations, theyobserved some very small relations among some pairs of scales (e.g., nei-ther Agreeableness nor Neuroticism was correlated with any of the sixHolland types), moderate relations among some pairs of scales (e.g.,Conscientiousness with Conventional interests, Extraversion with Enter-prising and Social interests), and some moderate-to-substantial correla-tions (Openness with Investigative, Artistic and Social interests). Basedon these analyses they highlighted four trait complexes across the threetrait domains: social, clerical/conventional, science/math, and intellec-tual/cultural. They suggest that abilities, interests, and personality developin tandem, such that the ability level and personality dispositions deter-mine the probability of success in a particular task domain, and interestsdetermine the motivation to attempt the task.

Sullivan and Hansen (2004) found that associations between person-ality traits and vocational interests can be accounted for by associationsbetween lower-order personality traits and interests. For example, theyfound the following associations: the trait of Warmth (a facet of Extraver-sion) largely accounted for the association between and Social interests


and Extraversion; Assertiveness (a facet of Extraversion) largely explainedthe association between Enterprising interests and Extraversion; and,Aesthetics (a facet of Openness to Experience) largely accounted forthe association between Artistic interests and Openness. Consistent withthe conclusion of Ackerman and Heggestad (1997), Sullivan and Hansen(2004) view interests as representing individual differences in the types ofactivities that individuals find enjoyable and motivating, whereas person-ality reflects behavioral tendencies, which play a role in both motivationand determining success in particular activities.

Two meta-analytic studies examined the nature and magnitude ofthe bivariate correlations among the five-factor model (FFM) personal-ity dimensions and Big Six Interests (Barrick, Mount, & Gupta, 2003;Larson, Rottinghaus, & Borgen, 2002). Although the two studies investi-gated the correlations between each of the Big Six and Big Five attributes,the results have important implications for understanding the higher-orderdimensions as well. The Barrick et al. (2003) study was based on 21 stud-ies, yielding 41 independent samples with a total sample size of 11,559.The Larson et al. (2002) meta-analysis was based on 24 samples with atotal sample size of 4,923. Only studies that used the NEO-PI or its short-ened version, the FFI, (Costa & McCrae, 1992) were included in the Larsonet al. study. Although the instruments and samples differed somewhat, bothmeta-analyses yielded similar results. For example, of the 30 possible pairsof correlations, only four were greater than or equal to .25 in the Barricket al. study (based on true score correlations): Investigative–Openness,Artistic–Openness, Social–Extraversion, and Enterprising–Extraversion.Similarly, only 5 of the possible 30 correlations were .25 or greater inthe Larson et al. study (based on uncorrected correlations). They were thesame four listed above for the Barrick et al. study, plus Conventional–Conscientiousness. Thus, the conclusion from these studies is that thereare moderate correlations between some personality traits and some inter-est types, whereas in other cases the correlations are very small. Further,the results show that of the correlations greater than .25, most represent re-lationships between interests and Factor β traits rather than Factor α traits.Collectively, this suggests that a higher-order dimension will emerge fromthe joint relationship of the RIASEC and FFM attributes that differentiateinterests from personality traits.

Hypothesis 1: A higher-order dimension will emerge that distinguishesinterests from personality traits. It will be defined by RIASEC interests atone end and by Factor α personality traits at the other end, with Factor βtraits falling in between.

This dimension is hypothesized to reflect fundamental differencesamong attribute types represented by personality traits and interest types.


However, the possible existence of this dimension does not preclude thepossibility that other higher-order dimensions further explain relations be-tween personality and interest domains. The additional dimensions maycapture fundamental motives showing how interests and personality traitsfunction jointly to influence behavior. Clearly, the dimensions identifiedby other researchers within the RIASEC types (e.g., Sociability and Con-formity) and within the FFM traits (Factors α and β) provide a solid foun-dation for conceptualizing what these dimensions could be. An importantissue that has not been addressed in the literature is whether these higher-order dimensions will emerge when both domains are considered jointly,and whether other meaningful dimensions will emerge as alternatives orin addition to them.

Based on the literature reviewed above and deductive reasoning, webelieve that two additional higher-order dimensions will explain relationsbetween RIASEC interest types and FFM dimensions. We conceptual-ize these two dimensions as representing broad-based motives that directand energize behavior at a basic level. Because the two dimensions ofConformity and Sociability identified for interests by Hogan (1982) arecompatible with FFM trait language, we derive the hypotheses using thosedimensions. Both the Larson et al. (2002) and Barrick et al. (2003) meta-analyses showed moderate correlations between Conventional interestsand Conscientiousness (r = .25 and ρ = .19, respectively). Thus, weexpect these two attributes to form a cluster that defines one end of the di-mension. The Larson et al. (2002) and Barrick et al. (2003) meta-analysesalso showed moderately strong correlations between Artistic interests andOpenness personality traits (r = .48, ρ = .39, respectively). Further,Digman’s (1997) research has shown a moderately strong correlation be-tween Extraversion and Openness personality traits. This suggests thatExtraversion and Openness personality traits and Artistic interests willform a cluster that defines the other end of a dimension. This dimension issimilar to the Conformity dimension proposed by Hogan (1982). However,given that it is defined jointly by interests and personality traits, whereasHogan’s analysis focused exclusively on interests, other interpretationsmay be warranted.

Hypothesis 2: A dimension will emerge that underlies RIASEC types andFFM interests that represent the broad motive of Conformity at one end andNonconformity at the other. It will be defined by both RIASEC interestsand FFM personality traits.

The Larson et al. (2002) and Barrick et al. (2003) meta-analyses alsoshowed moderately strong correlations between Social interests and Ex-traversion personality traits (r = .31, ρ = .29, respectively), and Enter-prising interests and Extraversion personality traits (r = .41, ρ = .41,


respectively). In addition, Holland’s circumplex predicts strong correla-tions between Social and Enterprising interests. We believe that theseattributes will cluster together to form one end of a dimension simi-lar to the Sociability dimension proposed by Hogan (1983). Holland’scircumplex shows that Realistic interests are the opposite of Social in-terests consequently, we expect that Investigative and Realistic interestswill cluster together to form the other end of this dimension. These at-tributes represent the preference to work alone, and with data and things.We do not have an expectation regarding which FFM traits would de-fine this end of the dimension. Collectively, this leads to the followinghypothesis.

Hypothesis 3: A dimension will emerge that represents preferences forinteracting with people at one end and preferences for interacting with dataand things at the other. It will be defined by both RIASEC interests andFFM personality traits.

In summary, we propose that the framework of the nomological net-work of RIASEC interest types and FFM personality traits consists ofthree higher-order dimensions. The first dimension differentiates the at-tributes according to two types of attributes, interests versus personalitytraits. The second two dimensions show that certain personality traits andinterests jointly influence two broad-based motives, the desire for confor-mity versus nonconformity, and desire for interacting with people versusinteracting with things.

Method

Meta-Analysis of FFM Traits and RIASEC Types

In order to examine the dimensionality of the RIASEC and FFM at-tributes, we formulated an 11 × 11 matrix consisting of true score cor-relations among the FFM dimensions and RIASEC types followed bymultidimensional scaling. The 11 × 11 matrix was formulated based onthe results of three meta-analyses: (a) correlations between the FFM traitsand RIASEC types, (b) intercorrelations among FFM types, and (c) in-tercorrelations among RIASEC types. The formulas developed by Hunterand Schmidt (1990) were used to conduct the meta-analyses. These pro-cedures yield estimates of population relationships after correction forvarious statistical artifacts. They also yield estimates of variability in theinter study results that may be attributed to statistical artifacts such assampling error, measurement error due to test unreliabilities, and rangerestriction.


Meta-Analysis of FFM-RIASEC Correlations

The correlations among the FFM traits and RIASEC types were ob-tained by updating the meta-analytic results obtained by Barrick et al.(2003) and Larson et al. (2002). As discussed earlier, both stud-ies examined the same basic questions, and yielded similar resultsand conclusions. For example, the five highest correlations between30 possible pairs of correlations between interest types and person-ality traits were the same in both studies: Investigative–Openness;Artistic–Openness; Social–Extraversion; Enterprising–Extraversion, andConventional–Conscientiousness. Generally speaking, the rank order ofthe correlations between the different personality–interest pairs was alsovery similar in both studies.

The scope of the Larson et al. (2002) study was narrower than Barricket al., as they restricted their studies to only one personality measure,the NEO-PI or the shortened version, the FFI (Costa & McCrae, 1992).The Barrick et al. study included multiple measures of personality, whichincreases the generalizability of the results, and their sample size wasmore than double that of Larson et al., which increases the robustnessof the findings. We examined the overlap between the two meta-analysesand found that there were four studies in Larson et al. (2002) that werenot included in Barrick et al. (2003). A search of the published literaturein the vocational psychology journals and personnel psychology journalsfor studies that had been published since the Barrick et al. study yieldedone additional study. Of the five studies, one did not report the necessaryeffect sizes and we were unable to obtain the other. Thus, we added threestudies to the Barrick et al. (2003) data, which yielded a total of five newindependent samples. This resulted in 46 studies that served as the basisof the updated meta-analysis.

The true score correlations among the FFM and RIASEC types arebased on 46 independent samples (k) with 12,433 individuals. A majorityof the samples used a Holland questionnaire (36 samples used the SDSor VPI) or a five-factor model personality questionnaire (31 samples).Furthermore, 22 of the 46 samples consisted of working adults, 16 con-sisted of college students, 4 consisted of military personnel, and 4 othersconsisted of both students and adults. We followed the criteria used byBarrick et al. to determine whether a study was included in the meta-analysis. The study had to report the minimum statistics necessary forconducting the meta-analyses such as zero-order correlations or the equiv-alent, and sample size. Further, the study had to include data on personalitymeasures that could be classified into the FFM personality factors, and usedoccupational measures that reported results using Holland’s typology, orhad to have a one-to-one correspondence with Holland’s six types.


Meta-Analysis of FFM Intercorrelations

The FFM intercorrelations were derived from a meta-analysis of cor-relation matrices from four frequently used personality inventories, eachof which was designed with the purpose of assessing the FFM person-ality traits (or some close variant of it). The four inventories were theNEO-PI (Costa and McCrae, 1992), the HPI (Hogan & Hogan, 1995), thePCI (Mount, Barrick, & Callans, 1999), and Goldberg’s personality inven-tory derived from the International Personality Item Pool (IPIP; Goldberg,1999)).

The NEO-PI-R is a measure of normal personality traits. It measuresthe FFM dimensions and six facets within each dimension. The normativesample used in this study was based on a composite of three subsam-ples: (a) a group of 405 men and women in the Augmented BaltimoreLongitudinal Study of Aging (ABLSA) who completed the instrument in1989 and 1990; (b) 329 ABLSA participants who completed the NEOPI-R by computer administration between 1989 and 1991; and (c) 1,539employees who participated in a national study of job performance. Toobtain a reasonably diverse normative sample, 500 men and 500 womenwere selected from these groups. They were first screened for validity andrandom responding and then selected to match the U.S. Census projec-tions for 1995. Coefficients alpha for the five scales were: Neuroticism .92,Extraversion .89, Openness .87, Agreeableness .86, and Conscientiousness.90. Thus the matrix for the NEO-PI was based on a normative sample of1000 adults.

The HPI assesses seven rather than five personality dimensions. Thedifference being that Extraversion is divided into two factors, Sociabilityand Ambition; in addition, there is a School Success scale (which was notused in this study). We used the average of the correlation between Socia-bility and Ambition with each of the five factors from the other inventoriesto represent the correlations with Extraversion. KR-20 coefficients for thescales were Adjustment .89, Ambition .86, Sociability .83, Likeability .71,Prudence .78, and Intellectance .78. The matrix for the HPI consisted ofa normative sample of 11,000 adults who were tested between 1984 and1992.

The matrix for the PCI consisted of a normative sample consistingof 4,140 adults (Mount et al., 1999). Coefficients alpha for the five scaleswere Neuroticism .92, Extraversion .89, Openness .87, Agreeableness .86,and Conscientiousness .90.

The matrix for Goldberg’s measure consisted of 486 adults. The IPIPis based on the lexically derived Big Five phenotypic model of personalityattributes (Saucier & Goldberg, 1998). Workers rated each item on a 5-point, Likert-type scale (from 1 = very inaccurate to 5 = very accurate).


The meta-analysis was conducted across a total of 10 pairs of vari-ables, each FFM personality factor with the other. The meta-analysis foreach pair of variables proceeded as follows. Although the sample sizes dif-fered across inventories, ranging from approximately 400 to 11,000, weweighted each inventory equally using a sample of 1,000. This avoidedthe potentially biasing effects associated with weighting a particular in-ventory more heavily than others. We then obtained the average observedcorrelations and the standard deviation (SDr) for this value. Each ob-served correlation was corrected for measurement error in both variablesusing the artifact distribution based on the reliabilities reported by the testauthors.

Meta-Analysis of RIASEC Correlations

In the vocational interest literature, a subset of 10 matrices has beenfrequently used as a normative sample. The use of these samples enhancesthe generalizability of structural analyses for the RIASEC types. Becauseof the possible nonindependence of the RIASEC matrices and the un-equal representation of the RIASEC measures, two matrices per measure(one based on a female sample and one on a male sample) were cho-sen to represent the variety of measures used to assess the RIASEC types.When possible, the correlation matrix came from a normative sample to en-hance the generalizability of the structural analyses. The 10 RIASEC sam-ples include (a) VPI seventh revision matrices (Gottfredson, Holland, &Holland, 1978) based on accidental samples of employed adults and col-lege students (378 females and 354 males); (b) SDS 1985 edition ma-trices (Holland, 1987) based on diverse, accidental samples, aged 14 to74 years, used to revise the SDS (470 females and 297 males); (c) SII-GOT matrices (Hansen & Campbell, 1985) based on the women- andmen-in-general samples: matching female (N = 300) and male (N = 300)samples (by occupational title) of employed adults, stratified according tolevel and Holland occupational classification, and randomly selected fromsamples used to construct the occupational scales; (d) CAI-enhanced ma-trices (Johansson, 1986) based on accidental samples of students and adults(703 females and 488 males); and (e) UNIACT matrices based on a na-tional probability sample of 12th grade students (2,218 males and 2,427 fe-males) that represents U.S. grade 12 enrollment in 1993 (American CollegeTesting, 1995).

The meta-analysis for each pair of interest types followed the procedurefor the personality measures described above. The sample sizes differedacross inventories, ranging from 400 to 700, and consequently we chose toweight each one equally using a sample of 400. This avoids the potentiallybiasing effects associated with a particular inventory. We then obtained the


average observed correlations and the standard deviation (SDr) for thisvalue. Each observed correlation was corrected for measurement error inboth variables using the artifact distribution based on reliabilities reportedby the test authors.

Cluster Analyses

Using the 11 × 11 matrix of true score correlations, we conducted acluster analysis to identify groups based on similarities among personality(Big Five) and vocational preference (Big Six) attributes. Although thereare several clustering methods such as single linkage, complete linkage,within-group averaging, and Ward’s minimum variance method, a reviewof simulation research on clustering methods (Milligan & Cooper, 1987)reported that the complete linkage method was better at recovering clus-ters than were the group-averaging or single linkage methods. Thus, weused the complete linkage method, which identifies clusters by joiningcases that are at the same level of similarity as all members of an existingcluster. The 11 attributes are grouped into a series of hierarchically relatedclusters in which members of each cluster are as similar to each other aspossible and members of distinct clusters are as heterogeneous as pos-sible. The resulting hierarchy can be represented by a dendrogram (treestructure). The PROC CLUSTER (Stat User’s Guide, 1990) was used toconduct the cluster analyses.

Multidimensional Scaling Analyses

We also conducted nonmetric multidimensional scaling (MDS) to ex-amine the relationships among the 11 variables using the 11 × 11 truescore correlation matrix. We used SAS PROC MDS (Stat User’s Guide,1990) to conduct nonmetric MDS analyses. MDS is typically based onmeasures of dissimilarity between variables but can also be based on cor-relations, covariances, and other measures of similarity (Davison, 1985;Kruskal & Wish, 1978). MDS is a class of techniques that are useful forinvestigating the structure of data. Using data indicating the degree ofsimilarity or dissimilarity between measures of various constructs, MDSattempts to locate points (representing each of the variables) in space insuch a way that the proximities between points are monotonically re-lated to the similarities between them. That is, points representing con-structs that are most similar will be located closest together, and thosethat are less similar will be located farther apart. Although MDS has notbeen used as extensively as factor analysis, there are circumstances inwhich MDS is the superior approach. One is in the analysis of correlationmatrices that have circumplex structures (Kluger & Tikochinsky, 2001).


Because empirical evidence supports the existence of Holland’s theo-rized circumplex of vocational interests, MDS is an appropriate method ofanalysis.

Property Vector Fitting

The technique of property vector fitting (Kruskal & Wish, 1978) wasused to assist the process of labeling the dimensions created in the MDSanalysis. The first step in property vector fitting is to obtain the set of coor-dinates from the MDS analysis that represent the underlying structure. Thesecond step is to obtain scores on proposed characteristics or propertiesfor each of the six interests and five personality attributes that may explainrelationships among interests and personality traits. To do this we createda list of 16 characteristics or properties that have been used in the researchliterature to define relationships among interests and/or personality traits.These 16 characteristics were: (a) Data—preference for working with data-related tasks, for example, keeping records, organizing files and numericdata; (b) Ideas—preferences for working with ideas, for example, cognitivetasks such as thinking, creative mental activity, using knowledge and in-sight; (c) Things—preferences for working with things, for example, work-ing with impersonal tasks such as dealing with machines, materials, andtools; (d) People—preferences for working with people, for example, inter-personal tasks such as thinking about others, taking care of others, teachingand directing others; (e) Conformity—preferences for conformity, for ex-ample, following rules, having a routine, enjoying structure, and have pre-scribed procedures; (f) Broad preferences—likes and dislikes for specifictypes of activities; (g) Personal style—enduring emotional, interpersonal,experiential, attitudinal, and/or motivational styles; (h) Reputation—howa person is perceived by others? (i) Identity—a person’s hopes, fears, am-bitions, and dreams; (j) Socialization—incorporating the norms of society,to restrain impulses, and/or reduce angry hostility; (k) Personal growth—desire to engage in personal growth, enlargement of self, and/or self-actualization; (l) Accomplishment striving—desire to accomplish things,for example, to get things done in a timely, careful, efficient way; (m) Sta-tus striving—desire to get ahead, for example, to prevail or to gain socialor material status in groups; (n) Communion striving—desire to get alongwith others, for example, to be accepted by the group, and the desire to pro-mote and maintain harmony in groups; (o) Abstractness—where abstractmeans the domain of the attributes is not well defined, and concrete meansthe domain of the attribute is very well defined; (p) Performance-orientedmotivation—engaging in goal setting, self-efficacy, and expectancybeliefs.


Nine PhD students who were enrolled in an OB/HR program and hadcompleted coursework in which Holland’s theory and the FFM person-ality traits were discussed, and one faculty member (not associated withthis study) served as raters. The 10 raters were presented with a packetthat consisted of definitions of the 11 attributes (6 interests and 5 per-sonality traits) and were asked to make ratings on a 7-point scale foreach attribute on the 16 characteristics or properties. The next step isto use linear multiple regression to regress the score for each propertyderived from the average of the 10 raters’ ratings onto the coordinatesobtained from the MDS analysis (discussed below) for the 11 attributesin order to assess the extent to which the property fits into the underly-ing structure. The salience of the structure is assessed by the varianceaccounted for (based on R2). Characteristics that achieved a significantR ( p < .05) were included in the figures that are discussed below. Theproperty vector is the line emerging from the origin of the MDS coordi-nate system that points in the direction of the strongest association betweenthe characteristic and the attribute. By examining the location of the dif-ferent property vectors in the interest structure, it is possible to identifythe structure among these attributes.

Results and Discussion

Table 1 shows the meta-analytic results for the intercorrelations amongthe RIASEC attributes based on the 10 RIASEC samples. Most of the truescore correlations are moderate in magnitude, with 11 of the 15 correla-tions greater than or equal to ρ = .20. We compared the average correlationbetween interest types that are adjacent to each other in Holland’s circum-plex to the average correlation of those that are not adjacent to each other.Consistent with Holland’s theory, the correlations were higher (ρ = .42)for adjacent types than those farther apart on the circumplex (ρ = .17).

Table 2 shows the meta-analytic results for the intercorrelations amongthe FFM personality traits. Of the 10 correlations, four were moderatelylarge, that is, exceeded ρ = .30. Three pairs of correlations are repre-sentative of Factor α: Stability–Conscientiousness (ρ = .52), Stability–Agreeableness (ρ = .42), Conscientiousness–Agreeableness (ρ = .39).The other correlation is consistent with Factor β, Extraversion–Openness(ρ = .45).

Table 3 displays the 11 × 11 true score correlation matrix consist-ing of the Big Six interests and the Big Five personality traits, whichwas derived from the three meta-analyses. Correlations among RIASECtypes and among FFM traits were obtained from the two meta-analysesdescribed above. The other correlations in the matrix were obtainedfrom the updated meta-analysis based on Barrick et al. (2003). Of the


TABLE 1Meta-Analytic Results for RIASEC Intercorrelations

Correlate ρ SDρ 90% CV % Var.

Realistic–Investigative 0.45 0.11 (.31, .60) 14.1Realistic–Artistic 0.25 0.06 (.17, .33) 40.1Realistic–Social 0.18 0.10 (.05, .31) 20.8Realistic–Enterprising 0.20 0.07 (.11, .30) 34.1Realistic–Conventional 0.27 0.10 (.14, .40) 20.5Investigative–Artistic 0.36 0.11 (.23, .50) 16.7Investigative–Social 0.26 0.15 (.07, .44) 6.14Investigative–Enterprising 0.09 0.12 (−.06, .23) 8.1Investigative–Conventional 0.17 0.11 (.04, .31) 11.4Artistic–Social 0.39 0.14 (.21, .56) 7.6Artistic–Enterprising 0.28 .18 (.13, .43) 16.1Artistic–Conventional 0.01 .13 (−.15, .18) 13.9Social–Enterprising 0.51 0.10 (.39, .63) .18Social–Conventional 0.29 0.13 (.12, .46) 13.0Enterprising–Conventional 0.53 0.13 (.36, .68) 10.6

Note. Number of samples in the analysis = 10; Total number of respondentsacross the samples = 3,760; ρ = estimated true score correlation (corrected forsampling error and unreliability); SDρ = estimated true standard deviation forthe correlation; 90% CV = estimated 90% credibility value for true score cor-relation; % Var. = percent variance in correlations accounted for by statisticalartifacts.

TABLE 2Meta-Analytic Results for Big Five Intercorrelations

Correlate ρ SDρ 90% CV % Var.

Stability–Extraversion 0.24 0.02 (.22, .27) 72.0Stability–Openness 0.19 0.12 (−.03, .34) 7.9Stability–Conscientiousness 0.52 0.19 (.27, .73) 2.9Stability–Agreeableness 0.42 0.13 (.26, .58) 7.8Extraversion–Openness 0.45 0.00 (.45, .45) 368.1Extraversion–Conscientiousness 0.17 0.11 (.04, .31) 10.5Extraversion–Agreeableness 0.26 0.15 (.07, .44) 6.14Openness–Conscientiousness 0.09 0.12 (−.06, .23) 8.1Openness–Agreeableness 0.17 0.11 (.04, .31) 11.4Conscientiousness–Agreeableness 0.39 0.14 (.21, .56) 7.6

Note. Number of samples in the analysis = 4; Total number of respondents across thesamples = 4,000; ρ = estimated true score correlation (corrected for sampling error andunreliability); SDρ = estimated true standard deviation for the correlation; 90% CV =estimated 90% credibility value for true score correlation; % Var. = percent variance incorrelations accounted for by statistical artifacts.


TABLE 3True Score Correlations Among RIASEC and FFM Variables

Variable 1 2 3 4 5 6 7 8 9 10 11

1 Realistic –2 Investigative 45 –3 Artistic 25 36 –4 Social 18 26 39 –5 Enterprising 20 09 28 51 –6 Conventional 27 17 01 29 53 –7 Conscientiousness 05 09 −06 07 08 19 –8 Agreeableness 01 01 03 17 −06 −01 39 –9 Openness 06 25 41 13 05 −10 09 17 –

10 Extraversion 03 02 09 29 40 06 17 26 45 –11 Emotional Stability 08 12 –02 04 08 03 52 42 19 24 –

Note. True score intercorrelations among the Big Five personality traits and amongthe Big Six interests are based on the meta-analytic results presented in Tables 1 and 2.Decimal points omitted.

30 possible true score FFM-RIASEC correlations (ρ), only 4 were greaterthan .20: Extraversion–Enterprising (ρ = .40), Openness to Experience–Artistic (ρ = .41), Extraversion–Social (ρ = .29), and Openness toExperience–Investigative (ρ = .25). Of the remaining correlations, 22were ρ = .10 or less. Thus there are moderate correlations between someinterest attributes and some personality attributes, but in general the corre-lations between interests and personality are low. As a general observation,correlations are highest for interest measures with other interest measuresand for personality measures with other personality measures.

Cluster Analysis

The hierarchical structure resulting from cluster analysis is shownby the dendrogram in Figure 1. Considering the RIASEC interests first,three clusters emerged in the following order: Enterprising–Conventional,Realistic–Investigative, Artistic–Social. These clusters correspond toGati’s (1991) three-group partition model of the structure of interestswhereby RIASEC types collapse into three clusters ([E, C], [R, I], and[A, S]). The Realistic–Investigative and Artistic–Social clusters thenformed a higher-order cluster, and then combined with the Enterprising–Conventional cluster to form the overall RIASEC interest cluster.

With respect to the FFM traits, Conscientiousness and Emotional Sta-bility formed the first cluster and Openness and Extroversion formed thenext cluster. The latter cluster is Digman’s Factor β. Then Agreeablenessjoins Conscientiousness and Emotional Stability to form a higher-order


Attribute

Realist Invest Artist Social Enterpr Convent Conscie EmStab Agree Open Extrovr

0.0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1.0 1.1 1.2 1.3 1.4

Figure 1: Cluster Analysis Results.

cluster, which is Digman’s Factor α. Then the Factor α and Factor β clus-ters join to form an overall personality cluster. The interest cluster and thepersonality clusters then combine to form the final cluster. Overall, theseresults are consistent with Gati’s three-group partition model of vocationalinterests, and support Digman’s two-factor structure of personality traits,α and β. In general the results of the cluster analysis show that interestsand personality attributes are distinct categories of attributes, as none ofthe interest types clusters with any of the personality traits until the finalstep. The results also show which of the Big Six interests and which ofthe Big Five personality traits are most similar to each other.

Spatial Analysis

Determining the appropriate dimensionality of an MDS solution ismuch like that of factor analysis in that statistical measures must be usedin conjunction with substantive considerations. We used Kruskal’s (1964)index of stress as well as interpretability to guide our choice of appropri-ate dimensionality. Stress is considered to be a “badness-of-fit” index withlarger values indicating a poorer solution. R2 values, indicating the corre-spondence between similarities and Euclidean distances between pointsin the solution, were used as well.


Dimension 1

+2 +1 0 -1 -2

-2 -1 0 +1 +2

Art

Inv

Soc

Real

Ent

Conv

Open Ext

Es Cons

Agr

Dimension 2

Figure 2: MDS Results for Dimensions 1 and 2.

We conducted a series of two-, three-, and four-dimensional MDSanalyses on the 11 × 11 RIASEC-FFM matrix of variables. The stressand R2 indices for the three models were as follows: two-dimensionalmodel (stress = .16, R2 = .81), three-dimensional model (stress = .06,R2 = .94), and four-dimensional model (stress = .04, R2 = .99). The three-dimensional model fit the data well and provided a substantially better fitthan the two-dimensional model. Although there was slight improvementfrom the three-dimensional model to the four-dimensional model, inspec-tion of the coordinates revealed that the four-dimensional model was noteasily interpretable. Further, Kruskal and Wish (1978) caution that whenthe ratio of objects to dimensions is much less than 4:1 (as is the case inthe four-dimensional solution), there is increased risk of capitalizing onchance. Thus, we focused our attention on the interpretation of the three-dimensional model. Because of the difficulties in displaying a model inthree-dimensional space, we present the results in two, two-dimensionalfigures. Figure 2 shows Dimensions 1 and 2 and Figure 3 shows Dimen-sions 2 and 3.

To facilitate the interpretation of the dimensions, we display the resultsof the vector fitting analysis in Table 4. The loadings for the vectors forthe 16 properties across the three dimensions and the R2 values are shown.


-2 -1 0 +1

+2

-2

-1

0

+1

+2

Dimension 3 Open

ArtAgr

ES

Con

Soc

ExtEnt

Conv

Real Inv

Dimension 2

Figure 3: MDS Results for Dimensions 2 and 3.

TABLE 4Property Vector Fitting Result

Rating scale attributes Dim 1 Dim 2 Dim 3 R2 p-value

1. Data .11 .82∗ .19 .71 .032. Ideas .24 −.76∗ .19 .71 .033. Things .33 .64∗ .62∗ .80 .014. People −.11 −.34 −.85∗ .84 .015. Conformity −.12 .84∗ −.10 .75 .016. Likes & dislikes .96∗ −.09 −.01 .94 .00017. Styles −.87∗ −.26 −.21 .91 .00058. Reputation −.85∗ −.23 −.19 .86 .0029. Identity .13 −.19 −.27 .11 .83

10. Incorporating norms −.78∗ .54 −.05 .91 .000411. Personal growth −.15 −.70∗ −.11 .53 .1312. Accomplish −.13 .80∗ .13 .66 .0413. Getting ahead −.03 −.08 −.66∗ .44 .2314. Getting along −.39 .03 −.37 .34 .3815. Abstract .23 −.80∗ .17 .72 .0216. Performance-oriented motivation −.21 .59 −.12 .57 .10

Note. Dim = Dimension; Values for Dimensions are Standardized beta-weights.∗p < less than .05.


Dimension 1

+2 +1 0 -1 -2

-2 -1 0 +1 +2

Art

Inv

Soc

Real

Ent

Conv

Open Ext

Es Cons

Agr

Likes and Dislikes

Personality

Reputation Factor αααα

Dimension 2

Figure 4: Vector Fitting Results for Dimensions 1 and 2.

A total of 11 of the 16 properties achieved significant R values ( p < .05).We used these 11 characteristics to aid in the interpretation of the threedimensions. Figures 4 and 5 show the results of the vector fitting analyses.

Dimension 1: Interests versus Personality Traits

With respect to Dimension 1, Figure 2 shows that the 11 attributescomprise three distinct clusters, RIASEC interests, and two personalityclusters. The six interest types appear in the top half of the dimensionand are clearly distinct from the five personality traits, which occupy thebottom part of the dimension. With respect to the FFM traits Conscien-tiousness, Emotional Stability, and Agreeableness form one cluster, andOpenness and Extraversion form the other. These results are consistentwith Digman’s (1997) Factors α and β and provide a type of validitycheck by showing that our meta-analytically derived FFM intercorrela-tions correspond to previous theory and research, even when derived froma larger set of individual difference attributes, which includes interests.Consistent with Digman’s terminology, and for ease of communication,we call the cluster of interest attributes Factor γ , to differentiate it fromFactors α and β.


Figure 4 shows the results of the vector fitting analyses for Dimen-sion 1. The six interest attributes cluster tightly together at the top ofDimension 1. This suggests that the variance shared by the six attributesis the desire to engage in activities that are liked and to avoid those thatare disliked. As would be expected, the vector that clearly defines the topend of this dimension was the attribute category associated with the def-inition of interests “. . . attributes that refer to a set of likes and dislikesfor specific types of activities.” The five personality traits cluster at thebottom end of Dimension 1. The vector that clearly defines the bottom ofthis dimension is personal style, which was associated with the definitionof personality traits by McCrae and Costa (1989), “enduring emotional,interpersonal, experiential, attitudinal, and/or motivational styles.” Thevector fitting results in Figure 4 also show that this dimension was alsoshown by Hogan’s (1982) conceptualization that personality traits definea person’s reputation, that is, how a person is perceived by others. Thereis also a vector that bisects Emotional Stability and Conscientiousness,which is consistent with Digman’s Factor α. Overall, the findings con-firm Hypothesis 1 as the first dimension shows a clear separation of the11 attributes into two domains. The RIASEC interests cluster together atone end of the dimension and represent broad preferences for engagingin specific types of reinforcing environments or for specific activities thatindividuals enjoy. The FFM attributes cluster together at the opposite endand collectively represent enduring personal styles that influence how in-dividuals interact with the environment or activities that they have chosen.Within the cluster of FFM traits, the position of Factor β traits betweenRIASEC interests and Factor α traits suggests that the motivational prop-erties of Factor β traits are more similar to interests than are Factor α

traits.A question raised by the results for Dimension 1 is what distinguishes

the six interest attributes from the five personality traits? As discussedearlier, interests and personality traits are similar in that they both influenceindividuals’ motivation, for example, choices people make about whichactivities to engage in, how much effort to exert, and how long to persistwith tasks. One possible substantive interpretation of Dimension 1 is thatit differentiates the attributes according to how they motivate behavior. Asshown in Figure 2, the six interest types, referred to earlier as Factor γ ,cluster tightly together. This means that regardless of content, all interesttypes share something in common and these commonalities are not sharedby personality traits. The variance that is shared among all interests isthe desire to engage in activities that are liked. When individuals are inenvironments that are congruent with their interests, they are more likely tobe happy because their beliefs, values, interests, and attitudes are supportedand reinforced by people who are similar to them (Dawis, 1991; Hogan


& Blake, 1999). Thus, interests motivate people to engage in activities andtasks that they like and to avoid those they dislike.

At the opposite end of Dimension 1 are Factor α traits, which pertainto impulse restraint, conscience, reduction of hostility and aggression,and neurotic defense. They trigger motivational processes associated withself-regulation—attentional, emotional, and effortful control. The meta-analytic results of Judge and Ilies (2002) showed that Conscientiousnessand Emotional Stability, two of the Factor α traits, were related to all threeforms of performance-oriented motivation, goal-setting, self-efficacy, andexpectancy beliefs. These traits are instrumental to an individual’s successat work and in life in general, as they influence the motivation to get alongwith others, willingness to incorporate the norms of society, and desire tobe a productive, functioning member of society (Digman, 1997).

One final issue pertaining to the interpretation of Dimension 1 is thepossibility that it reflects method variance associated with the types ofmeasures used. If true, this suggests that one explanation for the resultsof the cluster analysis and the results for Dimension 1 of the MDS isthat interests cluster together because they are measured using interestinventories, and that personality traits cluster together because they aremeasured using personality inventories. However, for several reasons webelieve this explanation is incomplete. First, we deliberately used multipleinterest inventories (SDS, VPI, UNIACT, SCII, CAI) and multiple person-ality instruments (Hogan, PCI, NEO-PI, and Goldberg’s IPIP) to mitigatethis possibility. Moreover, several substantive findings are inconsistentwith this idea. MDS results show that although γ interests cluster tightlytogether, the personality traits do not. Two personality clusters emerged,Factors α and β, as distinct categories of personality traits, yet the fivetraits were measured by the same personality instruments. Such results areunlikely if the relationships were due only to method effects. However,because we cannot rule out this explanation completely, we acknowledgethat Dimension 1 might represent an artifact or method factor.

Dimension 2: Striving for Personal Growth Versus Strivingfor Accomplishment

Figure 3 displays the relationships among the 11 attributes after con-trolling for differences in the two major types of attributes captured inDimension 1, for example, whether the attributes are interests or personal-ity traits. Taken together, Dimensions 2 and 3 clearly illustrate Holland’smodel, as the RIASEC circumplex is visibly shown in the figure. Al-though verifying the circumplex was not a major purpose of our study,the fact that it was clearly reproduced provides a validity check that the10 interest samples we used accurately capture Holland’s theory. The fact


Ideas

Abstract

-2 -1 0 +1

+2

-2

-1

0

+1

+2

Dimension 3 Open

ArtAg

ES

Con

Soc

ExtEnt

Conv

Real Inv

Dimension 2

Things

Data

People

Accomplishment

Conformity

Figure 5: Vector Fitting Results for Dimensions 2 and 3.

that the circumplex was reproduced when analyzed jointly with person-ality traits illustrates the robustness of the model. The figure also showsthe relationship of the FFM personality traits with the RIASEC model.Generally speaking, the personality traits occupy the middle portion ofthe circumplex and bisect it, thereby creating two distinct clusters ofinterests.

Both ends of Dimension 2 were defined by interest attributes and per-sonality traits. The right side was defined primarily by Conventional in-terests and Conscientiousness personality traits and to a lesser degree byEnterprising and Realistic interests. The left side was defined primarily byOpenness and Extraversion traits and Artistic interests. Figure 5 displaysthe results of the vector fitting analysis for Dimensions 2 and 3. Threevectors defined the right end of Dimension 2: desire to accomplish things(getting things done in a timely, careful, efficient way), conformity (fol-lowing rules, having a routine, and preferring structure), and preferencesfor working with data (keeping records, organizing files, manipulatingnumeric data). The vector that defined the left end of Dimension 2 wasthe preference for working with ideas (cognitive activities such as think-ing, creative mental activity, and knowledge and insights) and abstraction(attributes that are not well defined). Taken together, these results are con-sistent with the hypothesis that this dimension reflects Conformity versusNonconformity.


However, upon further inspection there may be a more appropriateconceptualization of the dimension. Hogan’s view that this dimension re-flected Conformity versus Nonconformity was based exclusively on hisanalysis of interests—Conventional at one end and Artistic at the other. Theaddition of Conscientiousness traits at one end along with Conventionalinterests changes the meaning somewhat of the right side of Dimension 2.The two main components of Conscientiousness are Achievement andDependability. Conscientious people are hardworking, thorough, orga-nized, punctual, and dependable. For this reason, we believe that this endof the dimension reflects more than conformity, although this is certainlya component. Rather, these attributes reflect the desire and motivation toget things done; hence, we refer to this end of the dimension as Strivingfor Accomplishment.

Collectively, the traits that represent the other end of the dimension areDigman’s (1997) Factor β, with Artistic interests added. Digman (1997)pointed out that Factor β is a very broad concept that is difficult to define.It is closely aligned with the perspective of personal growth theorists suchas Rogers (1961) and Maslow (1950), and has also been associated withstatus (Hogan, 1982), power (McAdams, 1985), and agency (Wiggins,1991). Collectively, the three attributes reflect a dynamic style of inter-acting with the environment, venturesome encounters with life, creativeself-expression, an appreciation of aesthetics (e.g., Sullivan & Hansen,2004), and openness to new ideas and thoughts. Figure 5 shows that thevector that defines the left-hand side corresponds to the desire to work withideas, that is, to engage in creative mental processes using knowledge andinsight, and with abstractness, that is, attributes that have domains thatare ill-defined. Further, although the multiple R was not significant, thebeta for personal growth (e.g., desire to engage in personal growth, en-largement of self, and/or self-actualization) was significant. Overall, thissuggests that these attributes reflect more than nonconformity. Rather, webelieve that this end of Dimension 2 represents the broad motive associatedwith Striving for Personal Growth as it reflects seeking new experiences,venturesome encounters with life, creative expression, and openness tonew ideas.

Dimension 3: Working with People Versus Working with Things

As illustrated in Figure 3, Dimension 3 is defined at the top end byRealistic and Investigative interests and at the bottom end by Extraversionand Enterprising and Social interests. The common variance shared by theattributes that define the top end of Dimension 3 (Investigative and Real-istic interests) appears to be the preference to work alone. For example,individuals with these attributes prefer to engage in activities where they


can fix, build, or plant things (Realistic), or investigate, study, or analyze(Investigative). These activities are generally engaged in with little oppor-tunity or need for social interaction. Figure 5 shows that the vector that ismost strongly associated with this end of the dimension is the preferenceto work with things. It is more closely aligned with Realistic interests thanwith Investigative interests.

The common variance tapped into by the attributes that define the bot-tom end of Dimension 3 (Enterprising and Social interests, and Extraver-sion personality traits) is social interaction. The vector that most stronglydefines these attributes reflects the preference for working with people, asit bisects the space between Extraversion and Social interests. It shouldbe noted that although each attribute pertains to interactions with others,the specific type of social activities may vary, such as leading, persuading,helping, mentoring, and/or nurturing. This is similar to Prediger’s (1982)conceptualization that one dimension of Holland’s circumplex representsworking with people versus working with things and is also generallyconsistent with Hogan’s (1982) conceptualization of Sociability.

General Discussion

The most important contribution of this study is that it clarifies therelationships among personality traits and vocational interests. Previousresearch has identified the basic similarities and differences between theinterest and personality domains, but the precise nature of the underly-ing structural relations among the attributes has eluded researchers fordecades. Meta-analyses of these relations have been helpful, but they im-plicitly assume bivariate relationships among interests and personalitytraits. Our results show that the relationships cannot be explained ade-quately using only two dimensions. The MDS analyses, which account formultivariate relations among the attributes, showed that a two-dimensionalrepresentation of the relationships between interests and personality traitsfits poorly. Rather, the similarities and differences among attributes rep-resenting the two domains are better explained using three dimensions.Stated succinctly, personality and interests comprise three distinctly dif-ferent types of motivational constructs (Factors α, β, and γ ), which jointlyinfluence two fundamental motives.

There are several implications of our findings. First, all models thatseek to explain job performance contain components related to motivationand ability. One implication of our findings is the motivational compo-nent of these models should include the three distinct types of individualdifference attributes (Factors α, β, and γ ) because each may influencebehavior in unique ways. Further, models should include a componentthat accounts for the fundamental motives associated with Striving for


Self-Growth versus Accomplishment Strivings and Interacting with Peo-ple versus Interacting with Things. Incorporating these components intomodels of job performance will help researchers to more fully account forthe effects of behavior under volitional control.

The results also have practical implications for the development of newmeasures of noncognitive attributes. Dimensions 2 and 3 show that certainpersonality traits and interests can jointly influence the same fundamentalmotives. For example, Social Interests and Extraversion traits representpreferences for interacting with others. Both attributes direct effort andattention toward environments and activities that involve interactions withothers. Similarly, Artistic interests and Openness personality traits rep-resent the desire for personal growth. Together, they operate in similarways to direct people to seek new experiences, to express oneself in artis-tic and creative ways, to use one’s intellect, and to challenge the statusquo. This suggests that it may be useful to develop multifaceted measuresthat combine interests and personality traits for those sets of attributeswhere interests and personality traits have been found to cluster together(e.g., Openness personality traits and Artistic interests; Social interestsand Extraversion traits). Such compound trait complexes may predict mo-tivational and performance outcomes better than either attribute measuredseparately.

Another implication of the findings is that although interests and per-sonality traits can influence the same basic motives, they may differ in howthey do this. For example, even though Social interests and Extraversionpersonality traits influence the motive to interact with others, the motiva-tional process by which they do this is different. Social interests motivatepeople to seek environments that involve social activities. Extraversiondoes this too, but it also refers to a particular style of operating in that en-vironment. That is, extraverted people are sociable, active, energetic, bold,dominant, and status seeking. These traits go beyond describing what theperson prefers to do, that is, engage in social activities, by defining howthe person interacts with the environment. For example, the meta-analyticresults of Judge and Ilies (2002) showed that Extraversion is related tothe propensity to set goals. In contrast, we were unable to find any re-search that shows that people high in social interests are more likely toset goals than those low in social interests. In summary, although interestsand personality traits can jointly influence the same basic motives, themotivational processes by which they do so may differ.

The results suggest several areas for future research. First, the re-sults reported here are based on two specific sets of individual differencemodels—the RIASEC types and FFM traits. Although these models arewidely accepted in their respective fields, they are not the only ways toclassify these two sets of individual differences. For example, personality


has been organized at the highest levels using from one to seven factors(Saucier & Goldberg, 2003), and Gati (1991) has proposed a hierarchicalstructure for interests. However, because the Big Six and Big Five havebeen shown to be quite robust across different populations and because weused multiple instruments to measure interests and personality, we expectthat our results are generalizable. Nonetheless, different results might havebeen obtained if relations had been examined using different classifica-tion schemes. In addition, future research might also examine whether thesame underlying motives would be obtained when another set of noncog-nitive attributes, values, are also considered in addition to interests andpersonality.

Another area of future research is to consider nonlinear relationshipsamong interests and personality traits. Hogan, Hogan, and Roberts (1996)encourage researchers to consider nonlinear relations that might exist be-tween personality traits, and by extension, to interests. It is possible thatone’s standing on other traits and interests may influence the way in whicheach trait operates. To examine the influence of these “profiles,” we wouldneed to add information regarding the interactions among RIASEC andFFM variables. The nature of the meta-analytic data in this study restrictedus to examining the main effects. Witt, Burke, Barrick, and Mount (2002)illustrate the value of such a configural approach, as they found that per-sonality traits interact when predicting job performance. Specifically, theyshow that highly conscientious workers who lack interpersonal sensitivity(i.e., low on Agreeableness) are less effective, particularly in jobs requiringextensive cooperation or interaction with others. Although it is not pos-sible to examine nonlinear relationships with meta-analytic data, futureresearchers could examine whether personality traits and interest typesinteract to influence individuals’ behavior.

There are several aspects of the study that strengthen the contribu-tions. First, multiple measures of both interests and personality traits wereused to mitigate the potential idiosyncratic effects associated with specificinstruments. Further, the correlations analyzed in the study were basedon large, diverse samples derived by meta-analyses, thereby reducing theeffects of sampling error and measurement error. In addition, our resultsextend previous meta-analyses in this area because we obtained intercor-relations within the FFM attributes and within the RIASEC attributes,which allowed us to investigate the higher-order dimensions of personal-ity and interests. Finally, the use of multidimensional scaling showed thatrelationships among interests and personality traits cannot be adequatelyexplained by two dimensions. Although meta-analytic studies of the rela-tions between interests and personality have advanced our understanding,they implicitly assume that the underlying structure is two dimensional.This explains, at least partially, why the precise nature of the relationshipbetween interests and personality has eluded researchers for decades. This


study reveals that three rather than two dimensions best explain the un-derlying structure of interests and personality traits.

Summary

Personality traits and vocational interests are two major noncognitiveindividual difference domains in the field of psychology. Both play animportant role in understanding human behavior because they influencemotivation—choices individuals make about which tasks and activitiesto engage in, how much effort to exert on those tasks, and how long topersist on those tasks. Understanding the sources of common and uniquevariance among attributes that comprise these two domains provides amore complete understanding of basic human motives. Although the rela-tionship between personality and interests has been generally understood,the precise nature of the structural relationship has eluded researchers fordecades. Overall, results clarified the underlying structure among inter-ests and personality traits by showing the ways in which interests andpersonality traits are both similar and different.

REFERENCES

Ackerman PL, Heggestad ED. (1997). Intelligence, personality, and interests: Evidence foroverlapping traits. Psychological Bulletin, 121, 219–245.

Adler A. (1939). Social interest. New York: Putnam.Allport GW. (1937). Personality: A psychological interpretation. New York: Holt.American College Testing. (1995). Technical manual revised unisex edition of the ACT

interest inventory (UNIACT). Iowa City, IA.Assouline M, Meir EI. (1987). Meta-analysis of the relationship between congruence and

well-being measures. Journal of Vocational Behavior, 31, 319–332.Bakan D. (1966). The duality of human existence: Isolation and communion in Western

man. Boston: Beacon.Barrick MR, Mount MK, Gupta R. (2003). Meta-analysis of the relationship between the

five-factor model of personality and Holland’s occupational types. PERSONNEL PSY-CHOLOGY, 56, 45–74.

Berdie RF. (1944). Factors related to vocational interests. Psychological Bulletin, 41, 137–157.

Bingham WV. (1937). Aptitudes and aptitude testing. New York: Harper.Block J. (1995). A contrarian view of the five-factor approach to personality description.

Psychological Bulletin, 117, 187–215.Carter HD. (1944). Vocational interests and job orientation: A ten year review. Applied

Psychology Monographs, No. 2.Cole NS, Hanson GR. (1978). Impact of interest inventories on career choice. In Diamond EE

(Ed.), Issues of sex bias and sex fairness in career interest measurement. Washington,DC: National Institute of Education.

Costa PT, Jr, McCrae RR. (1992). Revised NEO personality inventory (NEO PI-R) and NEOfive-factor inventory (NEO FFI) professional manual. Odessa, FL: PsychologicalAssessment Resources.

Cronbach LJ, Meehl PE. (1955). Construct validity in psychological tests. PsychologicalBulletin, 52, 281–302.


Dawis RV. (1980). Measuring interests. New Directions in Testing and Measurement, 7,77–91.

Dawis RV. (1991). Vocational interests, values, and preferences. In Dunnette MD, HoughLM (Eds.), Handbook of industrial and organizational psychology (pp. 833–871).Palo Alto, CA: Consulting Psychologists Press.

Davison ML. (1985). Multidimensional scaling versus components analysis of test inter-correlations. Psychological Bulletin, 97, 94–105.

Day SX, Rounds J. (1997). “A little more than kin, and less than kind”: Basic interestsin vocational research and career counseling. Career Development Quarterly, 45,207–220.

Digman JM. (1990). Personality structure: Emergence of the five-factor model. AnnualReview of Psychology, 41, 417–440.

Digman JM. (1997). Higher-order factors of the Big Five. Journal of Personality and SocialPsychology, 73, 1246–1256.

Funder DC. (2001). The personality puzzle (2nd ed.). New York: Norton.Gati I. (1991). The structure of vocational interests. Psychological Bulletin, 109, 309–324.Goldberg LR. (1999). A broad-width, public domain, personality inventory measuring facets

of several five-factor models. In Mervielde I, Deary I, De Fruyt F, Ostendorf F (Eds.),Personality psychology in Europe (Vol. 7, pp. 7–28). Tillburg, Netherlands: TilbergUniversity Press.

Gottfredson GD, Holland JL, Holland JE. (1978). The seventh revision of the vocationalpreference inventory. JSAS Catalog of Selected Documents in Psychology, 8, 98.(Ms. No. 1783).

Hansen JC, Campbell DP. (1985). Manual for the SVIB-SCII (4th ed.). Palo Alto, CA:Consulting Psychologists Press.

Hogan R. (1982). A socioanalytic theory of personality. In Page MM (Ed.), Nebraskasymposium on motivation 1982: Personality—current theory and research (pp. 55–89). Lincoln, NE: University of Nebraska Press.

Hogan R. (1991). Personality and personality measurement. In Dunnette MD, Hough LM(Eds.), Handbook of industrial and organizational psychology (Vol. 2, 2nd ed.,pp. 327–396). Palo Alto, CA: Consulting Psychologists Press.

Hogan R, Blake R. (1999). John Holland’s vocational typology and personality theory.Journal of Vocational Behavior, 55, 41–56.

Hogan R, Hogan J. (1995). The Hogan personality inventory manual (2nd ed.). Tulsa, OK:Hogan Assessment Systems.

Hogan R, Hogan J, Roberts BW. (1996). Personality measurement and employment deci-sions: Questions and answers. American Psychologist, 51, 469–477.

Holland JL. (1959). A theory of vocational choice. Journal of Counseling Psychology, 6,35–45.

Holland JL. (1966). The psychology of vocational choice: A theory of personality types andmodel environments. Waltham, MA: Blaisdel.

Holland JL. (1973). Making vocational choices: A theory of careers. Englewood Cliffs, NJ:Prentice-Hall.

Holland JL. (1985). Making vocational choices: A theory of careers. Englewood Cliffs, NJ:Prentice-Hall.

Holland JL. (1987). 1987 manual supplement for the self-directed search. Odessa, FL:Psychological Assessment Resources.

Holland JL. (1997). Making vocational choices: A theory of vocational personalities andwork environments (3rd ed.). Odessa, FL: Psychological Assessment Resources.

Holland JL. (1999). Why interest inventories are also personality inventories. In SavickasM, Spokane A (Eds.), Vocational interests: Their meaning, measurement, and usein counseling (pp. 87–101). Palo Alto, CA: Davies-Black.


Hunter JE, Schmidt FL. (1990). Methods of meta-analysis: Correcting error and bias inresearch findings. Newbury Park, CA: Sage.

Johansson CB. (1986). Career assessment inventory: The enhanced version. Minneapolis,MN: National Computer Systems.

Judge TA, Ilies R. (2002). Relationship of personality to performance motivation: A meta-analytic review. Journal of Applied Psychology, 87, 797–807.

Kluger AN, Tikochinsky J. (2001). The error of accepting the “theoretical” null hypoth-esis: The rise, fall, and resurrection of commonsense hypotheses in psychology.Psychological Bulletin, 127, 408–423.

Kruskal JB. (1964). Nonmetric multidimensional scaling. Psychometrica, 29, 1–27.Kruskal JB, Wish M. (1978). Multidimensional scaling. Beverly Hills, CA: Sage.Kuder GF. (1977). Activity interests and occupational choice. Chicago: Science Research

Associates.Larson LM, Rottinghaus PJ, Borgen F. (2002). Meta-analyses of Big Six interests and Big

Five personality factors. Journal of Vocational Behavior, 61, 217–239.Maslow A. (1950). Self-actualizing people: A study of psychological health. In Personality

symposia: Symposium No. 1 on values (pp. 11–34). New York: Grune & Stratton.McCrae RR, Costa PT, Jr. (1989). Reinterpreting the Myers-Briggs type indicator from

the perspective of the five-factor model of personality. Journal of Personality, 57,17–40.

McAdams DP. (1985). Power, intimacy, and the life story. Psychological inquiries intoidentity. Homewood, IL: Dow-Jones-Irwin.

Milligan GW, Cooper MC. (1987). Methodology review: Clustering methods. AppliedPsychological Measurement, 11, 329–354.

Mount MK, Barrick MR, Callans M. (1999). Manual for the personal characteristics in-ventory Chicago: The Wonderlic Corporation.

Mount MK, Muchinsky PM. (1978). Person-environment congruence and employee jobsatisfaction. A test of Holland’s Theory. Journal of Vocational Behavior, 13, 84–100.

Prediger DJ. (1982). Dimensions of underlying Holland’s hexagon: Missing link betweeninterests and occupations. Journal of Vocational Behavior, 21, 259–287.

Prediger DJ, Vansickle TR. (1992). Locating occupations on Holland’s hexagon: BeyondRIASEC. Journal of Vocational Behavior, 40, 111–128.

Rank O. (1945). Will therapy and truth and reality. New York: Knopf.Rogers CR. (1961). On becoming a person. Boston: Houghton Miflin.Rounds J, Tracey TJ. (1993). Prediger’s dimensional representation of Holland’s RIASEC

circumplex. Journal of Applied Psychology, 78, 875–890.Saucier G, Goldberg LR. (2003). The structure of personality attributes. In Barrick MR,

Ryan AM (Eds.), Personality and work (pp. 1–29). San Francisco: Jossey-Bass.Saucier G, Goldberg LR. (1998). What is beyond the Big Five? Journal of Personality, 66,

495–524.Stat User’s Guide. (1990). Cary, NC: SAS Institute Inc.Savickas ML. (1999). The psychology of interests. In Savickas ML, Spokane AR (Eds.),

Vocational interests: Their meaning, measurement, and use in counseling (pp. 19–56). Palo Alto, CA: Davies-Black.

Spokane AR. (1985). A review of research on person-environment congruence in Holland’stheory of careers. Journal of Vocational Behavior, 26, 306–343.

Sullivan BA, Hansen JC. (2004). Mapping associations between interests and personality:Toward a conceptual understanding of individual differences in vocational behavior.Journal of Counseling Psychology, 51, 287–298.

Super DE. (1973). The work values inventory. In Zytowski DG (Ed.), Contemporary ap-proaches to interest measurement. Minneapolis: University of Minnesota Press.


Strong EK, Jr. (1960). An 18-year longitudinal report on interests. In Layton WL (Ed.),The strong vocational interest blank: Research and uses. Minneapolis: Universityof Minnesota Press.

Tracey TJ, Rounds J. (1993). Evaluating Holland’s and Gati’s vocational-interest models:A structural meta-analysis. Psychological Bulletin, 113, 229–243.

Wiggins JS. (1991). Agency and communion as conceptual coordinates for the under-standing and measurement of interpersonal behavior. In Grove WM, Cicchetti D(Eds.), Thinking clearly about psychology. Vol. 2: Personality and psychopathology(pp. 89–113). Minneapolis: University of Minnesota Press.

Witt LA, Burke L, Barrick MR, Mount MK. (2002). The interactive effects of conscien-tiousness and agreeableness on job performance. Journal of Applied Psychology, 87,164–169.